By James Mackintosh
Artificial intelligence has saved the stock market -- again.
Plenty of investors and even central bankers have watched the S&P 500 hit record highs and worried that share prices have lost touch with the reality of $100-a-barrel oil.
But if there's a problem, it's not only, or even mainly, a failure to recognize the fragility of the cease-fire in the Gulf.
Everything depends on whether the AI boom is a bubble.
Here are a few facts often missing from the discussion about how stocks have performed since the U.S. and Israel started bombing Iran at the end of February:
-- In the S&P, 118 stocks have fallen more than 10%, notably those now
facing higher costs for fuel, aluminum and other raw ingredients, or
those dependent on sales to customers hit particularly hard, such as
farmers. That compares with only 82, mostly AI-related, that are up more
than 10%.
-- Exclude the AI version of the Magnificent Seven stocks -- Broadcom
alongside Alphabet, Amazon, Apple, Meta, Microsoft and Nvidia -- and the
market value of the S&P is actually down. Put another way, these seven
are lifting the entire market.
-- The average U.S. stock has fallen almost as much as the MSCI All Country
World Index excluding the U.S., because more than half the S&P members
are down. Yet the U.S. index is up 4%, and the tech-heavy Nasdaq 8%,
because the biggest stocks are so big they more than offset the falls
elsewhere.
-- Half the S&P's sectors are down. Two of those that are up -- consumer
discretionary and communications services -- are, like the market,
dominated by a large AI company. A majority of the stocks in those
sectors fell.
What really matters then is the same thing everyone in markets has been debating for the past year or more: whether there's a bubble in AI.
Investors are betting on continued rapid data-center construction, helping the picks-and-shovels stocks such as Nvidia, even as they don't expect positive free cash flow from AI developers themselves until the end of the decade.
Signs of froth are everywhere in AI. The IPOs of Anthropic and OpenAI are expected to be the biggest ever, and investors are desperate to find ways into the stocks before they list.
Data-center construction is a meaningful driver of the economy. Investors have proved willing to buy anything with an AI label slapped on it, whether one of the endless Silicon Valley startups or, briefly, sneaker maker Allbirds and Trump-affiliated payments company-turned-crypto treasury Alt5 Sigma. Even Cisco, the dot-com darling, this year finally passed its March 2000 share-price high thanks to its sales to data centers. It only took 26 years.
Hope overcomes every obstacle: The risk that AI becomes a widely available commodity, the constant hallucinations, repeated business-model pivots by OpenAI, massive investment requirements, deep uncertainty about the technology, rising political opposition and doubts about customers' willingness to pay.
The pro-AI case is simple: This time is different. AI will have bigger effects, and more quickly, than past technologies that led to bubbles, from canals and railways through bicycles and electricity to the dot-com bubble. It will transform the economy. Profits for AI companies will be mahoosive. And anyway, the leading stocks are still cheap compared with the dot-com bubble peak.
Ultimately we only ever know a bubble for sure after it bursts. If a technology lives up to the market's expectations, prices were right and it wasn't a bubble. But there are a lot of major questions the AI industry can't yet answer, and investors seem happy to believe it will all work out.
One basic point of history to remember: The biggest winner from the internet -- Google, now Alphabet -- wasn't even public when the bubble burst in 2000. Likewise, given uncertainty about the technology behind AI, there's no guarantee that today's leaders will ultimately come out on top.
Back then, the telecommunications companies were the picks-and-shovels suppliers of the fiber and home and office connections that made the internet run, and were investing as much capital as they could secure. Hopes were high, just as they are for the chip makers and other data-center suppliers raking in profits today. Back then, the hopes proved entirely unfounded.
Still, if the fantastical-sounding human-equivalent AIs are developed, this time could be different. Or, at least, the bubble could inflate a lot further before it pops. That's the problem with bubbles: Even if you correctly identify one, being early is the same as being wrong.
Write to James Mackintosh at james.mackintosh@wsj.com
(END) Dow Jones Newswires
April 26, 2026 01:32 ET (05:32 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments